Overeducation, skill mismatches, and labor market outcomes for college graduates

Overeducation and overskilling

Overeducation and overskilling are distinct phenomena, as indicated by their low correlation (r = 0.197). A novel feature of the new work based on HILDA panel data that is discussed here is the categorization of graduates into four mutually exclusive groups: well matched in both education and skills; overeducated only; overskilled only; and both overeducated and overskilled [1], [2]. This categorization proves to be very illuminating when the effects on wages, job satisfaction, and job mobility are compared (Figure 1). Thus, 28.4% of the sample are mismatched, earning less than matched workers and having lower job satisfaction if overskilled and a higher voluntary quit rate.

The panel data

Panel data estimation has distinct advantages over cross-section analysis for examining the impacts of job mismatch over time. Using methodologies based on panel data strongly reduces the size of many of the coefficients, throwing doubt on the results of cross-sectional analyses, which have predominated in the literature. The analysis using the Australian panel data shows that the relationship between job mismatch and labor market outcomes is strongly influenced by unobserved individual differences. Above all, the results shown in Figure 1 suggest that, at least at a descriptive level, it is a combination of overeducation and overskilling, rather than either one individually, that has the most damaging outcomes.

Although the HILDA data are for a single country and a group of male graduates, there is reason to believe that the Australian data are not that atypical. True, comparing 2005 HILDA data with data from the European Working Conditions Survey of the same year for all employees finds that the pattern of Australian educational and skill mismatch does not match that apparent across much of the EU, either in the aggregate or in individual countries, but there is also substantial variation across European counties in the extent of mismatch, and the Australia data are bounded by individual country estimates within the EU. For economy of space, the analysis here is limited largely to men. While the mismatch pattern is similar for women, the negative wage and job satisfaction effects are stronger for women than for men, which suggests that mismatch is more damaging for women.

The HILDA sample is restricted to an unbalanced panel of working-age, full-time, paid male employees with a college degree or equivalent qualification for whom complete information was available in the data set on the variables of interest such as wages, job satisfaction, and job mobility [1], [2]. The sample size is approximately 700 individual observations per wave. Overskilling is derived from HILDA data using the response on a seven-point scale to the statement “I use many of my skills and abilities in my current job.” The overeducation measure is obtained by comparing educational attainment with the modal level in each occupational group. Respondents are also asked to rank their job satisfaction on a scale of 0 to 10 [1], [2].

The results are highlighted here less for the findings themselves and more as illustrations of how to address shortcomings in the literature and emphasize important elements of the analysis. To allow for comparisons with the earlier literature, a pooled ordinary least squares (OLS) model was used to estimate the wage effects of mismatch as a benchmark. Pooling across waves enables a direct comparison to be made with cross-section studies. For job satisfaction and job mobility, which involved binary variables, a pooled probit model was estimated as a benchmark (see Econometric models). For dealing with unobserved individual heterogeneity, a random effects probit model with a Mundlak correction was estimated.

Wage penalties

Controlling for unobserved individual differences removes most of the negative wage impact found in studies with cross-section data for male college graduates who are overeducated only or overskilled only (Figure 2). Only male graduates who change from a well-matched job to one for which they are both overeducated and overskilled suffer a wage penalty compared to the omitted reference category (of just under 6%, in both random effects and fixed effects models, as opposed to over 30% in the OLS estimates).

Job satisfaction

Job satisfaction is treated as an outcome of mismatch by observing the impact that each type of mismatch has on the level of satisfaction after controlling for other factors that might influence satisfaction (Figure 3). Where a mismatch does not reduce job satisfaction, it is likely that the mismatch reflects a voluntary underutilization of qualifications or skills. Being overeducated on its own has no discernable effect on job satisfaction among the HILDA sample. (A similar result was obtained for the UK [3].) In contrast, being overskilled, whether alone or combined with overeducation, greatly reduces job satisfaction. Thus, while cross-section analysis shows a negative marginal effect on job satisfaction of being overskilled of 22.2%, panel estimation reduces the effect to only 6.9%. For people who are both overskilled and overeducated, the reduction in job satisfaction is much larger, at 15.2%.

Job mobility/quitting

Only overeducation on its own, as shown by the significant coefficient for voluntary quits when panel estimation is used, or jointly with overskilling, as shown by the significant 0.593 coefficient, increases the probability of quitting as a consequence of job mismatching (Figure 4). Male college graduates who are overskilled only are no more likely to quit than are male graduates who are well matched, after the analysis controls for unobserved individual differences, as indicated by the –0.066 insignificant coefficient. This finding, when combined with the negative effects of mismatch on job satisfaction as shown in Figure 3, suggests that overeducated graduates or those who are both overeducated and overskilled tend to be trapped in jobs that have undesirable characteristics.

Persistence of the effects of job mismatches

Some studies have suggested that overeducation may be a form of investment in training or experience that can boost future returns to human capital [4]. If true, this would mean that mismatch was a temporary phenomenon, which would greatly reduce the need for policy intervention. To determine whether that is the case, studies need to examine the propensity of mismatched workers to remain in mismatched jobs.

Persistence can be examined using a dynamic random effects probit specification that includes a lagged dependent variable. Using this approach with HILDA data, and incorporating nine waves covering the period 2001–2009, shows that an overskilling mismatch is highly persistent in a manner that is inversely related to education level [2]. The effect of previous overskill mismatches on present overskilling mismatches is positive but diminishes over time. The first lag of an overskilling mismatch increases the present probability of an overskilling mismatch for a college graduate by 15 percentage points, and so on across waves. A college graduate who has never been overskilled for a job has a 4.6% probability of becoming overskilled in the following year. By contrast, a college graduate who was overskilled in each of the previous three years has a 38% probability of being overskilled in the following year. Although data are available for only a limited period, this suggests high persistence or a scarring effect from being in an overskill mismatch.

Though college graduates suffer less from persistence than do other mismatched workers, they suffer a greater wage penalty than any other group. The results also suggest that higher paid college graduates face the largest wage losses, consistent with higher wages being offered as compensation for taking greater risks. Public policy therefore needs to consider not only the extent and persistence of skill mismatch, but also the size and persistence of the associated wage effects of such persistence if the problem is to be targeted efficiently.

How results with pooled data differ from those with cross-section data

The results reported here differ from those of studies based mainly on cross-section data in a number of ways. First, the results for the pooled data show a significant wage penalty only for male college graduates who are both overskilled and overeducated but not for those who are only overskilled or only overeducated. Second, overeducation on its own has no adverse effect on job satisfaction. Job satisfaction is affected only by overskilling on its own or combined with overeducation. Third, unobserved individual heterogeneity matters. In several cases, the panel estimations reduce the significance of findings compared with those using cross-section data. Together, these differences suggest that policy attention should focus on preventing overskilling, particularly when combined with overeducation, rather than on overeducation alone. Job recruitment should aim to secure a better match between skills of new hires and the jobs they fill. Doing so could benefit both employees and employers by boosting productivity, improving morale, and reducing the quit rate [5].